An eddy covariance set-up for methane measurements

69

4444

A compact and stable eddy covariance set-up

for methane measurements

using off-axis integrated cavity output spectroscopy

This chapter was published as Hendriks, D.M.D., Dolman, A.J., Van der Molen, M.K.,

Van Huissteden, J., 2008. A compact and stable eddy covariance set-up for methane

measurements using off-axis integrated cavity output spectroscopy. Atmos. Chem. Phys.

8, 431-443.

Abstract

A Fast Methane Analyzer (FMA) is assessed for its applicability in a closed path eddy

covariance field set-up in a peat meadow. The FMA uses off-axis integrated cavity output

spectroscopy combined with a highly specific narrow band laser for the detection of CH4

and strongly reflective mirrors to obtain a laser path length of 2-20×103 m. Statistical

testing and a calibration experiment showed high precision (7.8×10-3

ppb) and accuracy (<

0.30%) of the instrument, while no drift was observed. The instrument response time was

determined to be 0.10 s. In the field set-up, the FMA is attached to a scroll pump and

combined with a 3-axis ultrasonic anemometer and an open path infrared gas analyzer for

measurements of carbon dioxide and water vapour. The power-spectra and co-spectra of

the instruments were satisfactory for 10Hz sampling rates.

Due to erroneous measurements, spikes and periods of low turbulence the data series

consisted for 26% of gaps. Observed CH4 fluxes consisted mainly of emission, showed a

diurnal cycle, but were rather variable over time. The average CH4 emission was 29.7

nmol m-2

s-1

, while the typical maximum CH4 emission was approximately 80.0 nmol m-2

s-1

and the typical minimum flux was approximately 0.0 nmol m-2

s-1

. The correspondence

of the measurements with flux chamber measurements in the footprint was good and the

observed CH4 emission rates were comparable with eddy covariance CH4 measurements

in other peat areas.

Additionally, three measurement techniques with lower sampling frequencies were

simulated, which might give the possibility to measure CH4 fluxes without an external

pump and save energy. Disjunct eddy covariance appeared to be the most reliable

substitute for 10Hz eddy covariance, while relaxed eddy accumulation gave reliable

estimates of the fluxes over periods in the order of days or weeks.

Chapter 4

70

4.1 Introduction

Methane is the third most important greenhouse gas globally, after water vapour and

carbon dioxide. Its concentration has risen by 150% since the pre-industrial era (Foster et

al., 2007) and currently 20% of the enhanced greenhouse effect is considered to be due to

methane (Foster et al., 2007). Although methane is less abundant in the atmosphere

compared to carbon dioxide, it is a relatively strong greenhouse gas: the Global Warming

Potential of methane expressed as CO2-equivalents is approximately 25 (over 100 years)

while that of carbon dioxide is by definition 1 (Foster et al., 2007). For water vapour and

carbon dioxide, the eddy covariance technique has been used for several decennia

(Aubinet et al., 2000) to measure the exchange of these gases with the earth surface. In

eddy covariance, the vertical flux (Fs) of an atmospheric property (s) is directly

determined by the covariance of that property and the vertical velocity, as shown in Eq.

(4.1). This can be obtained by calculating the time averaged product (over the period t1 to

t2) of the deviation (s’) of the atmospheric property (s), from 'sss += , and the deviation

(w’) of the vertical wind velocity (w) from 'www += :

∫−

==

2

112

)(')('1

''

t

t

s dttstwtt

swF (4.1)

(Aubinet et al., 2000). The eddy covariance technique requires an instrument with high

precision, accuracy and system stability as well as high sampling rates and short

instrument response time (τ). Unfortunately, the low abundance of methane in the

atmosphere hampers adequate concentration measurements of this gas and the eddy

covariance technique has therefore been rarely applied for the assessment of methane

emissions (Kroon et al., 2007; Hargreaves et al., 2001; Kormann et al., 2001). The

advantages of the eddy covariance technique compared to other techniques for measuring

trace gases are nonetheless obvious: integrated continuous measurements over a large

footprint area (102 to 10

4 m

2) and longer periods without disturbance from small scale

surface features. These properties enable assessments of spatial and temporal variability at

the landscape scale.

Since the 1980s, infrared absorption spectrometry using tunable diode lasers (TDLs) has

been widely used for measurements of trace gases in the lab. A field technique for eddy

covariance using a multipass absorption cell was introduced in 1995 (Zahniser et al.

1995). Unfortunately, serious problems of drift and low sensitivity effects occurred.

Additionally, the method had more practical drawbacks: the large nitrogen Dewar for

temperature control that needs refill weekly, an extensive and sensitive optical module,

and frequent calibration. Improvements were made with the Quantum Cascade Laser

(QCL) spectrometer which was first introduced in 1994 (Faist et al., 1994). The method

was more stable and accurate in eddy covariance set-up than the TDL spectrometry

(Nelson et al., 2004; Kroon et al., 2007), but the practical drawbacks of the methane

instruments (large nitrogen Dewar, optical module, and frequent calibration) were still

present in the QCL spectrometer technique. Hitherto, the micrometeorological

An eddy covariance set-up for methane measurements

71

measurement techniques for methane have thus implied very expensive, large and labour

intensive field set-ups with logistic limitations to unattended field deployment.

The objective of this study was to investigate the applicability and quality of the FMA, a

new off-axis integrated cavity output spectroscopy (ICOS) technique using a highly

specific narrowband laser, for its applicability for eddy covariance field measurements of

methane. The relatively user-friendly and low cost set-up is tested for precision, accuracy

and system stability as well as its’ performance in eddy covariance field set-up. The first

data series are analysed and compared with data obtained by existing measurement

techniques. To asses the applicability of the instrument at locations that lack external

power supply, alternative measurements techniques of the set-up are simulated which

could reduce the power requirements.

4.2 Methods and materials

4.2.1 Instrument design

A measurement cell with highly reflective mirrors was combined with a highly specific

narrowband laser system by O’Keefe et al. (1998), creating a path length of at least two

km in the measurement cell. In this manner, small absorption rates cause much larger

reduction of the total transmitted laser intensity. The ICOS technique could therefore be

used for the detection of gases with ultra weak absorption, while making the extensive

optical module and the nitrogen Dewar superfluous. The operation of the DLT-100 Fast

Methane Analyzer (FMA) from Los Gatos Research Ltd. is based on this improved ICOS

technique and uses a distributed feedback (DFB) diode laser with a wavelength of nearly

1.65 µm. The DFB diode laser offers tunability, narrow line width and high output power

in a compact and very rugged setup. It features a grating structure within the semi-

conductor, which narrows the wavelength spectrum and guarantees single-frequency

emission. Off-axis ICOS implies that a laser beam directed into the measurement cell at a

slight angle, after which it is reflected in the cell numerous times by highly reflective

mirrors (reflectivity ~ 0.9999), thus creating a path length of 2-20×103 m by making 1-

10×104 passes in the cell (Bear et al., 2002; Fig. 4.1). The detector measures fractional

absorption of light at the methane resonant wavelength, which is an absolute measurement

of the methane concentration in the cell.

The path length of the laser, and therefore the time period during which the laser is being

reflected in the cell for each measurement, is dependent on the reflectivity of the mirrors

in the measurement cell. This period over which the laser is being reflected in the

measurement cell is called the mirror ringdown time (MRT) and is continuously

monitored by the FMA. The MRT cannot be allowed to drop below 3.0 to 3.5 µs, since the

laser path length then becomes too short to detect the changes in laser intensity properly.

The mirrors in the measurement cell are sensitive to dirt accumulating in the measurement

cell; a small contamination of the measurement cell causes rapid decrease of the MRT.

Cleaning the mirrors is a relatively simple procedure that can be done in a dust-free

environment.

Chapter 4

72

The FMA measures in the concentration range from 10 to 25×103 ppb and operates

autonomously. Technically, measurements can be made at rates up to 20Hz and at ambient

temperatures of 5 oC to 45

oC, while humidity should be below 95% to avoid

condensation. The pressure in the measurement cell (Pcell), which can be adjusted by a

valve switch on the instrument, should be kept near 210 hPa. To obtain sampling rates

higher then 1Hz, an external pump is needed to maintain the required Pcell and τ.

The measurement cell has a volume of 0.55×10-3

m3 and a length of 0.20 m (Fig. 4.1).

Data output is provided in analogue as well as digital format (RS232&TCP/IP) and the

device can store data up to 10 gigabytes. Warm-up time is approximately one minute and

measurements as well as performance can be observed on a colour TFT LCD flat panel

display. The dimensions of the FMA are 0.25 m height, 0.97 m width and 0.36 m depth

and it has a weight of 22 kg. Power requirements are 115/230 VAC, 50/60Hz and 180W

and inlet/outlet fittings are of the Swagelok type (3/8”,

1/4”).

4.2.2 Site description

Besides testing in the laboratory, the FMA was tested in an eddy covariance set-up at the

Horstermeer measurement site. This site is located in a eutrophic peat meadow area in the

central part of the Netherlands and was described extensively by Hendriks et al. (2007).

The area has a flat topography and vegetation consists of grasses, small shrubs and reeds.

Before, CH4 fluxes in the area were measured with the flux chamber technique and

variation between three land elements was observed: emissions from the saturated land

0.21 m

0.06 m

~0.50 m

0.20 m

Figure 4.1: Schematic overview of the integrated cavity output analysis (ICOS)

technique used in the FMA. (Source: Los Gatos Research Ltd.) “R” is the reflectivity

of the mirrors and “T” and “P” are the sensors of Tcell and Pcell.

An eddy covariance set-up for methane measurements

73

and water surfaces were high compared to the relatively dry land. The annual weighted

CH4 emission at the site over 2005 and 2006 was estimated at 83.95 ± 54.81 nmol m-2

s-1

(Hendriks at al., 2007).

4.2.3 Assessment of instrument stability, precision and accuracy

System stability is a major factor influencing high-sensitivity measurements.

Theoretically, the signal from a perfectly stable system could be averaged infinitely.

However, real systems are stable only for a limited time period. The length of time over

which a laser signal can be averaged to achieve optimum sensitivity, and thus high

precision, largely determines the quality of the spectrometer. The precision of

concentration measurements should be at least a few parts per thousand of the ambient

mixing ratio., which is approximately 4 ppb for a mixing ratio of approximately 1800 ppb

(outside air) in the case of CH4 (Kroon et al., 2007). Both maximum system stability and

precision can be determined using the Allan variance (Allan, 1966; Werle et al., 1993;

Nelson et al., 2004; Kroon et al., 2007). The Allan variance (2

Aσ ), as a function of

integration time Τ, is the average of the sample variance of two adjacent averages of time

series of data and is described by Eq. (4.2):

2

11

2)]()([

2

1)( kAkA

mk s

m

ss

tA −= ∑

=

+σ (4.2)

with: ∑=

++=

k

llkss x

kkA

1)1(

1)( , with: s = 1,…..,m and m = m’ – 1

I

n this equation A is an average of the CH4 concentration, k is the number of elements in

subgroup x, s is the subgroup number, l is the sample-number in the subgroup, and m’ the

number of independent measurements. It is assumed that data are collected over a constant

data interval ∆t, therefore the integration time Τ = k∆t (Allan, 1966; Werle et al., 1993;

Nelson et al., 2004). The Allan variance decreases when random noise dominates over

drift effects. However, when noise caused by instrumental drift of the system starts to

dominate, the Allan variance starts to increase, indicating a decrease of system stability

and hence precision. CH4 concentration measurements over a ten minute period of relative

constant CH4 concentrations with a mean value of 1905 ppb and a standard deviation of

4.74 ppb with 10Hz sampling rate were used to determine the Allan variance.

Subsequently, the Allan variance of this period was plotted over the integration time Τ

(Fig. 4.2), showing a decreasing Allan variance over integration times larger than 2.4 s.

No increase of the Allan variance was observed at larger integration times, indicating an

absence of instrumental drift, and thus high system stability and precision, for integration

times of a few seconds up to ten minutes. Additionally, an indication for the short term

precision (σ) was obtained by the y-axis intercept at the minimum Allan variance (2

Asσ =

6.1×10-3

ppb2). Using Eq. (4.3):

Chapter 4

74

2/1−×= sAs fσσ (4.3)

in which Asσ is the square root of the minimum Allan variance and fs is the sampling

frequency of the system (10Hz), σ was determined as 7.8×10-3

ppb Hz-1/2

. Precisions of

0.3 ppb Hz-1/2

(Nelson et al., 2004) and 2.9 ppb Hz-1/2

(Kroon et al., 2007) were found in

previous studies of QCL instruments.

Next, a calibration experiment was carried out by administering standard gas mixtures

with concentrations of 125 ppb and 2002 ppb CH4, both with an accuracy of 0.2 ppb, to

the FMA at ten instances within a 10-day period. During the whole experiment the FMA

was never turned off in order to imitate longer measurement periods as will be the case in

the field set-up. The deviation from the standard gas values of CH4 (MD) was calculated

for each measurement by Eq. (4.4):

1.82

1.86

1.90

1.94

1.98

0 100 200 300 400 500 600

time (sec)

CH

4 c

oncentr

ation (

ppm

)

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1 10 100 1000integration time (s)

Alla

n V

ariance б

A2 (ppm

2)

a.

b.

Figure 4.2: Time series of CH4 concentration measurements with 10Hz sampling rate with a

mean of 1905 ppb and a standard deviation of 4.74 ppb (a.) and the Allan variance plot for

these data (b.).

An eddy covariance set-up for methane measurements

75

MD = S

SM − (4.4)

where M is the measured CH4 concentration in ppb and S is the CH4 concentration of the

standard gas in ppb. A two-point calibration factor (Fcal) was calculated for each set of

measurements by Eq. (4.5):

Fcal =

LH

LH

MM

SS

−

− (4.5)

where SH and SL are the high and low standard values of CH4 in ppb and MH and ML are

the high and low measured values of CH4 in ppb. Although the measured concentration

sometimes varied one or two ppb from the standard values over the experimental period

for both gases, no actual drift was observed in the instrument (Table 4.1). Fcal was 1.000

on average with fluctuations < 0.30%, indicating high accuracy of the FMA.

Additionally, over a period of a week, CH4 concentration measurements at the

Horstermeer site were compared with CH4 concentrations measured at 20 and 60 m height

at a measurement site in Cabauw (51o58’N and 4

o 55’E)

1, approximately 30 km from the

Horstermeer site. The CH4 concentrations at Cabauw were measured with a Carlo Erba

1 The CH4 concentration measurements in Cabauw are measured with a gas chromatograph by the Energy

Centre of the Netherlands (ECN).

time

date time elapsed M L M H

(days) (ppb) (ppb)

02/04/2007 11:50:00 0.000 125 -0.10% 2005 0.20% 0.998

03/04/2007 09:00:00 0.882 127 1.50% 1999 -0.10% 1.003

03/04/2007 15:30:00 1.153 125 -0.10% 1999 -0.10% 1.002

04/04/2007 09:20:00 2.896 125 -0.10% 2000 -0.10% 1.001

04/04/2007 16:45:00 2.205 125 -0.10% 1999 -0.10% 1.002

05/04/2007 09:30:00 2.903 125 -0.10% 1999 -0.10% 1.002

05/04/2007 17:30:00 3.236 124 -0.90% 2001 0.00% 1.000

10/04/2007 09:45:00 7.913 125 -0.10% 2004 0.10% 0.999

11/04/2007 09:40:00 8.910 125 -0.10% 2002 0.00% 1.000

11/04/2007 16:30:00 9.194 125 -0.10% 2002 0.00% 1.000

12/04/2007 09:15:00 9.892 125 -0.10% 2004 0.10% 0.999

measurement

MD L MD H

Calibration experiment FMA

low CH4 gas (125 ppb) high CH4 gas (2002 ppb)F cal

Table 4.1: Results of the 10-day calibration experiment with two standard gases of 125 and

2002 ppb respectively. Time elapsed since start of experiment, CH4 concentrations measured

by FMA (ML and MH), MD values and Fcal values are shown.

Chapter 4

76

gas chromatograph system and have a precision of 2 ppb. The CH4 concentration at the

Horstermeer site was 15 ppb higher on average, which might be the result of the relatively

high CH4 emissions from the peat meadow area in which the measurements were taken.

Nonetheless, the increasing trend of CH4 concentration at the Horstermeer site was similar

to the trend at the Cabauw site. Generally, slow variation in concentrations is caused by

the difference between continental and marine background concentration, the continental

concentration being approximately 50 ppb higher (Eisma et al., 1994). During the

measurement period, prevailing winds were from the east (continental). This accounted for

the slow rise in CH4 concentrations at both measurement locations.

The influence of changing temperature and pressure conditions in the measurement cell on

CH4 concentration measurements was assessed in the laboratory. During a time series of

continuous measurements, a step change in Pcell was induced, while the temperature in the

measurement cell (Tcell) increased steadily. An effect of the increase of Tcell on CH4

concentration was observed neither from the time series, nor from the correlation plots

(Fig. 4.3). A decrease in the CH4 concentration data was observed 990 sec after the sharp

decrease in Pcel. However, this feature did not result in a clear correlation between CH4

1895

1900

1905

1910

1915

1920

1925

1930

0 0.5 1 1.5 2 2.5time (hours)

CH

4 c

on

c.

(pp

b)

31.2

31.6

32.0

32.4

32.8

0 0.5 1 1.5 2 2.5time (hours)

tem

pe

ratu

re (

oC

)

197

198

199

200

0 0.5 1 1.5 2 2.5

time (hours)

pre

ssu

re (

hP

a)

1895

1900

1905

1910

1915

1920

1925

197.0 197.5 198.0 198.5 199.0 199.5 200.0pressure (hPa)

CH

4 c

on

ce

ntr

atio

n (

pp

b)

1895

1900

1905

1910

1915

1920

1925

31 31.5 32 32.5 33

temperature (oC)

CH

4 c

on

ce

ntr

atio

n (

pp

b)

R2 = 0.00002

R2 = 0.0853

Figure 4.3: Results of the laboratory experiment on the effect of Tcell and Pcell on CH4 concentration

measurements. a.: time series of CH4 concentration measurements; b.: time series of Tcell

measurements; c.: time series of Pcell measurements; d.: scatter plot of CH4 concentration against Tcell

with correlation coefficient (R2); e.: scatter plot of CH4 concentration against Pcell with R

2.

An eddy covariance set-up for methane measurements

77

concentration and Pcell. Also, this type of Pcell changes did not occur under normal

circumstances when Pcell had a stable value near 210 hPa.

4.2.4 Incorporation of FMA in eddy covariance system

After testing in the laboratory, the FMA was installed in a closed-path eddy covariance

field set-up (Fig. 4.4). In order to obtain 10Hz measurements, a dry vacuum scroll pump

(XDS35i, BOC Edwards, Crawly, UK) was used with a maximum pumping speed of

9.72×10-3

m3 s

-1. However, at the required pressure of 210 hPa the actual pumping speed

was 5.50×10-3

m3 s

-1. The scroll pump has relatively high power requirements: 100/200V

to 120/230V, 50/60Hz and 600W. It was placed at the end of the set-up, connected to the

FMA by a wire-reinforced tube with an internal diameter of 1.9×10-3

m, sucking air

through the system. The FMA was placed in a heated, water resistant box, while the scroll

pump was placed in an aerated box that prevented it from getting wet and from

overheating. In addition to the internal filter with a pore size of 2 µm (Swagelok part no.

SS-4FW4-2), a filter with a pore size of 60 µm was placed at the inlet in order to prevent

dust, aerosols, insects and droplets from entering the tubing. The inlet was shielded from

the rain by a stainless steel cap. To prevent any water that accidentally passed the first

filter from moving down toward the analyzer the air was first led up through a stainless

steel tube (diameter of 6.4×10-3

m) that bends sharply at 0.5 m after which the air moves

data from

sonic

anemometer

to handheld

system

data from sonic to

handheld system

1+2

3

4

7

5

8 81011

12

15

direction data

direction air flow

3x25A;

17

14

6

9 16

13

1 rain cap (sharpened at the edges)

2 filter (60 µm)

3 bended tubing of stainless free steel

4 teflon tubing (diameter=64 mm)

5 water locked case

6 tube with warming strip

7 Fast Methane Analyser

8 heating elements

9 sturdy pipe to pump

10 semi closed case

11 scroll pomp

12 air exhaust with silencer

13 3-axis utra sonic anemometer

14 Licor7500 (CO2 and H2O analyser)

15 data collection in Gill software

16 handheld computer

17 power supply (regional electricity network)

swagelok screws are used for all connections

1.20 m

1.00 m 0.70

4.30

1 rain cap (sharpened at the edges)

2 filter (60 µm)

3 bended tubing of stainless steel

4 teflon tubing (diameter=64 mm)

5 water locked case

6 tube with warming strip

7 Fast Methane Analyser

8 heating elements

9 sturdy pipe to pump

10 semi closed case

11 scroll pomp

12 air exhaust with silencer

13 3-axis ultra sonic anemometer

14 Licor7500 (CO2 and H2O analyser)

15 data collection in Gill software

16 handheld computer

17 power supply (regional electricity network)

swagelok screws are used for all connections

Figure 4.4: Schematic overview of the combined field set-up of the closed path eddy

covariance system for CH4 using the FMA and the open path eddy covariance system

for CO2 and water vapour.

Chapter 4

78

down toward the analyzer through a Teflon tube (diameter of 6.4×10-3

m). After the air

has passed through the FMA measurement cell it flows through the scroll pump and was

exhausted through a silencer. The gas-inlet filter was positioned 0.2 m away from the LI-

7500 open path infrared gas analyzer (LI-COR Lincoln, NE, USA) and a Windmaster Pro

3-axis ultrasonic anemometer (GILL Instruments Limited, Hampshire, UK) directed into

the prevailing wind. Both instruments were installed at 4.3 m above the surface at the

Horstermeer measurement site (Hendriks et al., 2007). Data were logged digitally on a

handheld computer at a rate of 10 Hz (Van der Molen et al., 2006).

4.2.5 Assessment of FMA in measuring CH4 fluxes

In general a sampling rate of 10Hz, with a Nyquist frequency of 5Hz, is used for eddy

covariance techniques (Aubinet et al., 2000; Kroon et al. 2007). Therefore an instrumental

time response of 10Hz or greater is required in order to correlate with the wind

measurements made with the 3-axis ultrasonic anemometer. The measurement rate of an

instrument is determined by both electronic signal processing and by the τ of the

measurement cell (Nelson et al., 2004). Electronic signal processing is dependent on the

spectral complexity of the measurement technique as well as the technical design of the

ICOS technique. In the case of the FMA, this was defined by the designers (Los Gatos

Research Ltd.) as 20Hz. The limiting factor of the maximum sampling rate is often τ,

which could be determined by the volume of the measurement cell (0.55×10-3

m3) divided

by the actual pumping speed (5.50×10-3

m3 s

-1) giving a flow response of 0.10 s.

Additionally, τ was determined by applying a step change in concentrations at 20Hz

sampling rate (Fig. 4.5) (Moore, 1986; Zahniser et al., 1995; Nelson et al., 2004). Each

data point was the average mixing ratio of multiple step change events at a certain t (time

elapsed since step change in concentration). The τ was defined by the exponential fit to the

decay of the CH4 mixing ratio and was calculated τ = 0.11 s by Eq. (4.6):

[CH4]t = [CH4]t=0 e (-t/ τ)

(4.6)

0

0.2

0.4

0.6

0.8

1

-0.5 0.0 0.5 1.0 1.5 2.0

time (s)

dC

ob

s/d

Cin

cr

tube length = 0.25m

tube length = 1.00m

Figure 4.5: Averaged and

normalised time series of CH4

concentration data showing

the instrument response with

20Hz sampling rate changing

from ambient air to a gas

sample with a high CH4

concentration for tube lengths

of 0.25 m and 1.0 m.

An eddy covariance set-up for methane measurements

79

10-3

10-2

10-1

100

101

10-3

10-2

10-1

100

n [Hz]

n S

wC

H4

(n)

Observed wCH4

Theoretical w-scalar

10-3

10-2

10-1

100

101

10-3

10-2

10-1

100

n [Hz]

n S

w(n

)

Observed w

Theoretical w

10-3

10-2

10-1

100

10-3

10-2

10-1

n [Hz]

f S

w(n

)

Observed CH4

Theoretical scalar

a.

b.

c.

Figure 4.6: Averaged and binned

results of the spectral analyses of the

FMA eddy covariance set-up for two

moderately turbulent days (six half

hour periods per day). a.: Observed

and theoretical power-spectrum w;

b.: Observed and theoretical power-

spectrum of [CH4]; c.: Observed and

theoretical co-spectrum of w’[CH4]’.

Chapter 4

80

where [CH4]t is the CH4 concentration at instance t. In this, sensitivity to tube length could

be observed (Fig. 4.5). The data in the graph are normalized by the transformation

dCobs/dCincr, where dCobs is the increase of [CH4] between the starting time and time t and

dCincr is difference between the final [CH4] and [CH4] at the starting time. Since the effect

of tube damping was corrected separately, and τ only refers to the response of the

instrument itself, the actual τ was determined as 0.10 s.

The suitability of the FMA eddy covariance set-up was further evaluated by examining the

power spectra of w and [CH4] and co-spectrum of the covariance w’[CH4]’ (Stull et al.,

1988; Kaimal and Finnigan, 1994). For this purpose twelve half hours of data from two

days with moderately unstable conditions were analysed. The results of the spectral

analysis were averaged and binned and the logarithmic spectral and co-spectral densities

were plotted against frequency (Fig. 4.6). Additionally, the theoretical spectra defined by

Kaimal et al. (1972) and Højstrup (1981) were plotted in the graphs (Smeets et al., 1998).

The power spectrum of w very closely followed the whole theoretical spectrum, while

both the [CH4] power spectrum and the w’[CH4]’ co-spectrum showed slight deviations

from the theoretical spectrum. Most important is that the inertial subranges of both the

power spectra and the co-spectrum were in general agreement with the theoretical curve.

The [CH4] power spectrum showed a slight tendency to extend upward in the low

frequency range. This was also observed in the [CO2] power spectrum from the LI-7500

analyzer at the site (Hendriks et al., 2007), indicating that the upward tendency in the low

range was not instrument related, but was rather an effect of environmental factors.

Kaimal et al. (1976) found high densities of low frequencies in the temperature power

spectrum in response to a diurnal cycle. Since a diurnal cycle was observed for both [CO2]

and [CH4] at the Horstermeer site, this might have been the cause of the upward tendency

observed in the power spectra. However, the overestimation of low frequencies will

automatically cause an underestimation of the high frequencies, since the total area under

the power spectrum is a fixed surface (Kaimal et al., 1972; Højstrup, 1981). Considering

the w’[CH4]’ co-spectrum, the observed peak near n=10-1

Hz generally matched the peak

of the theoretical spectrum and was in agreement with spectral analysis of CH4 flux

measurements by Verma et al. (1992), Kormann et al. (2001) and Werle and Kormann

(2001). The spectral performance of the eddy covariance set-up for temperature and CO2

fluxes were discussed previously in a paper by Hendriks et al. (2007) and were in

agreement with the spectral analyses of the CH4 flux measurements.

Additionally, in previous research by Hendriks et al. (2007) the energy balance consisting

of the eddy covariance measurements of latent heat and sensible heat, the micro-

meteorological measurements of incoming and reflected radiation components and the

ground heat flux showed a closure of 82%. This indicated that the data from the eddy

covariance set-up were acceptable (Lloyd et al., 1997).

The travel time of the air in the closed path set-up from the inlet filter to the FMA, caused

a time lag with respect to the in situ measured wind data. For 20 half hour periods, the

covariance w’[CH4]’ of w’ at instance t = 0.0 s and [CH4]’ at t = 0.0 s was determined, as

well as the covariance of w’ at t = 0.0 s and [CH4]’ at the t = 0.1 s, t = 0.2 s, t = 0.3 s, …, t

= 2.0 s. For all half hour periods the highest value of w’[CH4]’ occurred with [CH4]’ at t =

An eddy covariance set-up for methane measurements

81

0.6 s. For the calculation of the actual covariance the time lag of the CH4 measurements

compared to the wind measurements was therefore taken as 0.6 s.

4.2.6 Data processing

The EUROFLUX methodology (Aubinet et al., 2000) was applied to the eddy covariance

data to calculate the CH4 fluxes from the closed path system and the CO2 fluxes from the

open path system (Hendriks et al., 2007) on a thirty minute basis. Since system

instrumental drift was not observed, an overestimation of the fluxes as a result of this

averaging period was not expected. The damping effect of the τ on the CH4 signal was

corrected for in the flux calculation procedure as well as for the time lag of 0.6 s between

closed path CH4 and open path wind, temperature and CO2 measurements (Moore et al.,

1986). The tube length of the set-up was over 1000 times the inner diameter of the tube

and therefore the air temperature in the measurement cell could be considered stable. As a

result, the Webb correction for density fluctuations arising from variations in temperature

that was applied to the open path CO2 measurements, was not required for the closed path

measurements of CH4 (Leuning and Moncrieff, 1990). The Webb correction for density

fluctuations arising from variations in water vapour (measured with the LI-7500) was

applied according to Leuning and Moncrieff (1990). Frequency loss corrections for

closed-path systems were applied according to the theory of Leuning and King (1992).

Since 3-axis ultrasonic anemometers were found to under measure wind speed at large

angles, the method of Nakai et al. (2006) was used to apply the angle of attack dependent

calibration (Gash and Dolman, 2003, Van der Molen et al., 2004).

4.2.7 Simulation of alternative flux measurement approaches

In eddy covariance, the sampling rate determines the number of samples that are taken out

of an infinitely large number of samples. The higher the sampling rate, the higher the

statistical reliability, which results in higher accuracy of the observed means and

covariances (Van der Molen, 2002). In order to obtain reliable estimates of fluxes, a

sampling rate of at least 10Hz is generally used for eddy covariance techniques. However,

measurements performed at lower rates than 10Hz can generate reliable results too (Rinne

et al., 2000 and 2001; Graus et al., 2006; Businger and Oncley, 1990). 1Hz eddy

covariance, disjunct eddy covariance, 1Hz eddy covariance and relaxed eddy

accumulation (REA) are possible alternatives for the preferred 10Hz eddy covariance.

These measurement techniques could make the external pump of the FMA eddy

covariance set-up superfluous and thus save over half of the energy required for the eddy

covariance set-up. This might be necessary for operation in remote places where an

external power source is not available. Here, the raw CH4, CO2 and water vapour 10Hz

field data were manipulated to simulate 1Hz eddy covariance, disjunct eddy covariance

and REA.

With 1Hz eddy covariance, Fs is determined in the same manner as 10Hz eddy covariance

(Eq. (4.2)), but at a lower frequency. The method was tested by averaging each 10

consecutive data points of CH4 concentration as well as wind velocity for the 10Hz data

set, thereby simulating a slower sampling rate (1Hz) with a longer τ (1.0 s instead of 0.10

s).

Chapter 4

82

Disjunct eddy covariance uses a subset of the whole 10 Hz time series to determine the

flux of an atmospheric property Fs according to Eq. (4.7):

∑=

×>==<

N

iiis sw

NswF

1

''1

'' (4.7)

where N is the subset of the data (Rinne et al., 2000 and 2001). Here, the disjunct eddy

covariance method was tested by sampling the first data point of every 10 data points from

the 10Hz data set, thereby creating a time interval of 0.9 s between the sampling moments,

while the τ remained 0.10 s. To build a field set-up for disjunct 1Hz eddy covariance a

‘snap sampling’ instrument would have to be mounted in front of the inlet of the system to

obtain samples that are sampled with a frequency of 1Hz, while maintaining the sampling

duration as short as possible (‘snap’). This method would give the FMA an analysis time

of 1.0 s instead of 0.10 s.

REA is a conditional sampling method in which air samples are drawn into two separate

reservoirs depending on the direction of w. The criterion of valve switching is based on

values of the standard deviation of w ( wσ ), which is measured by the 3-axis ultrasonic

anemometer at 10Hz. The valve is activated according to the threshold

condition ww σ6.00 = . In the case of 00 www ≤≤− (deadband values), neither “up”

nor “down” samples are taken, but samples are discarded from the sampling system

(Graus et al., 2006). Here, we simulated REA by dividing all eddy covariance data points

of one half hour measurement period into three data matrices based on the direction of the

w (upward, downward and deadband values). Next, the measured concentrations are

summed and the turbulent flux of the scalar s (Fs) was determined according to Eq. (4.8):

sbF ws ∆××≈ σ (4.8)

(Businger and Oncley, 1990). The b-value is the correction for the deadband and s∆ the

difference between the concentrations in the accumulation reservoirs (Bowling et al.,

1999).

4.3 Results

4.3.1 First data series

From the eddy covariance data series 11% consisted of gaps due to failure of the eddy

covariance set-up caused by rain events and instrumental errors. Additionally, 3% of the

data series consisted of spikes (CH4 flux > 100.0 nmol m-2

s-1

and CH4 flux < -10.0 nmol

m-2

s-1

) and were removed from the data set. During nighttime periods with low friction

velocity (u*) the turbulence of the atmosphere can become too low for the performance of

eddy covariance measurements (Wohlfahrt et al., 2005; Dolman et al., 2004). In order to

determine the critical u* value for CH4 eddy covariance measurements at this specific site,

the CH4 flux data and u* data from periods with incoming shortwave radiation (SWin) <

An eddy covariance set-up for methane measurements

83

20W m-2

were selected (Dolman et al., 2004). The nightly CH4 fluxes showed a

significant decrease for periods with u* < 0.09 m s-1

(Fig. 4.7). This result is similar to the

critical u* value of 0.10 m s-1

found for CO2 fluxes at the same site (Hendriks et al., 2007).

CH4 flux data measured during periods with u* < 0.09 m s-1

occurred at 12% of the data

series and where removed too. The total amount of data gaps accounted for 26% of the

whole data series.

CH4 fluxes ánd CO2 fluxes (net ecosystem exchange (NEE)) were plotted for a two week

period in June 2006 at the Horstermeer site (Fig. 4.8). Although the CH4 fluxes are rather

variable over time, a diurnal cycle can be observed with low emission during the night and

high emission during the day. CO2 fluxes have a similar, but opposite, diurnal cycle. The

observed CH4 fluxes consisted mainly of emission and had an average of 29.7 nmol m-2

s-

1, while the typical maximum CH4 emission was approximately 80.0 nmol m

-2 s

-1. The

typical minimum flux was approximately 0.0 nmol m-2

s-1

and at three occasions a small

uptake was observed.

4.3.2 Intercomparison with flux chamber data

For two days during the CH4 eddy covariance measurement period at the Horstermeer site,

flux chamber measurements were made in the footprint of the eddy covariance tower with

a Photoacoustic Field Gas-Monitor (type 1312, Innova AirTech Instruments, Ballerup,

Denmark) connected with tubes to closed, dark chambers (Hendriks et al., 2007). Flux

chamber data were collected at the three land elements occurring in the footprint of the

eddy covariance tower: relatively dry peat land, saturated peat land and ditch water

surfaces. The fluxes from the various land elements are averaged with respect to their rela

tive surface area (70%, 20% and 10% respectively; Hendriks et al., 2007). On June 10 the

-20

-15

-10

-5

0

5

10

15

20

25

0 0.1 0.2 0.3 0.4

u* (m s-1

)

w'[C

H4

]' (n

mo

l m

-2 s

-1)

Figure 4.7: Results of

the analyses of the

effect of low turbulence

on nightly CH4 fluxes,

showing a drop in flux

magnitude below of u*

of 0.09 m s-1

. Data are

binned over u* classes

of 0.02 m s-1

and error

bars show the standard

deviation per class.

Chapter 4

84

flux chamber measurements showed an average CH4 flux of 114.5 ± 9.6 nmol m-2

s-1

,

while the eddy covariance measurements showed an average CH4 flux of 83.2 nmol m-2

s-1

over the same period (Fig. 4.9). On October 3 the average CH4 flux from the chamber

measurements was 53.6 ± 11.2 nmol m-2

s-1

, while that from eddy covariance was 61.6

nmol m-2

s-1

.

4.3.3 Simulation of alternative flux measurement approaches

The simulated 1Hz and disjunct eddy covariance data as well as the simulated REA data

were compared with the original 10Hz eddy covariance data as half hourly averages over a

15 day period (Table 4.2). It can be observed that the various methods show similar but

not identical results and that all alternative methods showed variations from the 10Hz

eddy covariance data (Fig. 4.10). The 1Hz eddy covariance method showed on average a

slight overestimation of the half hourly fluxes for CH4 measurements (2%) and the

standard deviation of the time series was somewhat higher than that of the 10Hz eddy

covariance. Average CO2 and water vapour fluxes determined with 1Hz eddy covariance

showed however a large underestimation compared to the 10Hz eddy covariance measure-

-30

-20

-10

0

10

20

30

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191

time (julian days)

w'[C

O2]'

(µ

mol m

-2 s

-1)

-20

0

20

40

60

80

100

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191

time (julian days)

w'[C

H4]'

(nm

ol m

-2 s

-1)

a.

b.

Figure 4.8: CH4 flux data series over a two week period in the summer of 2006 (a.) and CO2

flux (NEE) data series over the same period (b.).

An eddy covariance set-up for methane measurements

85

ements. The disjunct eddy covariance method showed on average a slight underestimation

of the half hourly fluxes for CH4 as well as for CO2 and water vapour measurements (-2%

and -3%), while the standard deviation was somewhat higher than that of the 10Hz eddy

covariance.

The b-value used in REA for correction of the deadband area was determined by Pattey et

al. (1993) as 0.56. For the measurement set-up investigated here, a b-value of 0.20 showed

results most similar to the normal 10Hz eddy covariance measurements for CH4

measurements as well as for CO2 and water vapour measurements. The results of the REA

method showed on average a small underestimation of the half hourly fluxes for CH4 as

well as for CO2 and water vapour measurements (-4% and -2%). The standard deviation

however, was significantly higher than that of the 10Hz eddy covariance, pointing at

relatively large deviations from the 10Hz measurements at a half hourly basis.

0

50

100

150

200

250

300

350

400

450

10:00 11:00 12:00 13:00 14:00 15:00

time-of-day

w'[C

H4]'

flu

x (

nm

ol m

-2 s

-1)

EC datadry landsaturated landditch water surface

average of flux chamber data

July 10 2006

flux chamber

measurements

0

50

100

150

200

250

300

350

400

10:00 11:00 12:00 13:00 14:00 15:00

time-of-day

w'[C

H4]'

(n

mol m

-2 s

-1)

EC data

dry land

saturated landditch water surface

average of flux chamber data

October 3 2006flux chamber

measurements

Figure 4.9: Hourly CH4

flux data at July 10 and

October 3 (both 2006)

plotted in combination

with flux chamber data

from various land

elements in the footprint

of the eddy covariance

tower collected at the

same day. The square

marks show the weighed

average of the flux

chamber measurements.

Chapter 4

86

CH4 flux (nmol m-2

s-2

) 26.74 27.35 26.29 25.76

CO2 flux (µmol m-2

s-2

) -6.53 -5.84 -6.31 -6.41

H2O flux (mmol m-2

s-2

) 3.13 2.79 3.05 3.01

CH4 flux (%) 0% 2% -2% -4%

CO2 flux (%) 0% -11% -3% -2%

H2O flux (%) 0% -11% -3% -4%

CH4 flux (nmol m-2

s-2

) 11.65 12.23 12.7 14.34

CO2 flux (µmol m-2

s-2

) 9.95 9.13 9.86 11.22

H2O flux (mmol m-2

s-2

) 2.82 2.59 2.8 2.77

average gas flux

average deviation

from normal 10Hz

eddy covariance

standard deviation

10Hz EC 1Hz ECDisjunct

1Hz EC

REA

(d=0.6;b=0.2)

Table 4.2: Summary of results of 1Hz eddy covariance, disjunct eddy covariance and REA

compared to 10Hz eddy covariance for CH4, CO2 and water vapour (H2O) flux

measurements. Average gas flux over the 15 day period, average deviation from the 10Hz

eddy covariance data and standard deviations of half hourly data are shown.

-10

0

10

20

30

40

50

179.5 179.6 179.7 179.8 179.9 180.0 180.1 180.2 180.3 180.4 180.5

time (julian days)

w'[C

H4]'

(n

mo

l m

-2 s

-1)

10Hz EC

1HZ EC

disjunct 1Hz EC

REA (d=0.6 and b=0.20)

Figure 4.10: Time series of CH4 fluxes for a one day period: half hourly averages of 10Hz

eddy covariance and simulated 1Hz eddy covariance, disjunct eddy covariance, and REA.

An eddy covariance set-up for methane measurements

87

4.4 Conclusions and discussion

The FMA, which uses the new off-axis ICOS technique, was found to perform

satisfactorily in laboratory experiments and in the eddy covariance field set-up. Compared

to other techniques, the absence of a large nitrogen Dewar for cooling that would need

weekly refill, the compact, narrow band and stable laser, the relative user friendliness and

low costs are considerable advantages. The FMA eddy covariance set-up was found to

perform independently in an unattended field situation. However, care should be taken

when placing the instrument with a scroll pump in the field. The FMA should be kept dry

and in ambient temperatures between 5oC and 45

oC, while the scroll pump should be kept

dry and cool. Contamination of the measurement cell should be prevented, since the

mirrors inside the measurement cell are sensitive and only a small contamination might

cause a rapid decrease in reflectivity of the mirrors and performance of the instrument.

The analysis of the Allan variance indicated a high precision and system stability of the

FMA. Additionally, the calibration experiment showed sufficiently high accuracy (< 0.30

%). As long as ambient temperatures do not exceed the range of 5 to 45 oC and pressure in

the measurement cell is near 210 hPa, the CH4 measurements are not affected by changes

in Tcell and Pcell. Since no instrumental drift was observed in the Allan variance analyses, in

the calibration experiment or in the comparison with CH4 concentration data at a field site

nearby, it was concluded that frequent calibration of the FMA was not necessary.

The observed τ was 0.10 s, implying a maximum sampling rate of 10Hz which is

sufficient for eddy covariance measurements. The closed path FMA eddy covariance

system performed well, as shown by the power- and co-spectra which corresponded well

to the theoretical spectral curves. Additionally, the energy balance showed satisfactory

closure. CH4 fluxes appeared to be underestimated during periods with low turbulence and

a u* correction was applied for CH4 flux data during periods with u* < 0.09 m s-1

. Due to

erroneous measurements, spikes and periods of low turbulence the data series consisted

for 26% of gaps.

Observed CH4 fluxes consisted mainly of emission, showed a diurnal cycle, but were

rather variable over time. The average CH4 emission was 29.7 nmol m-2

s-1

, while the

typical maximum CH4 emission was approximately 80.0 nmol m-2

s-1

and the typical

minimum flux was approximately 0.0 nmol m-2

s-1

. These CH4 fluxes were in agreement

with QCL flux measurements at a managed peat meadow site in the Netherlands

(Reeuwijk), where emissions were 40 ± 31 nmol m-2

s-1

on average (Kroon et al., 2007)

and average annual CH4 emission of 30 nmol m-2

s-1

from peat lands in Germany and the

Netherlands (Drösler et al., 2007). From the comparison of the eddy covariance

measurements with flux chamber measurements, it was observed that the fluxes from the

two techniques showed a discrepancy of approximately 20%. However, considering the

fact that the flux chamber measurements are point measurements at the soil surface while

the eddy covariance has a footprint of hundreds of square meters, some degree of variation

may be expected.

Additionally, raw CH4, CO2 and water vapour 10Hz field data were manipulated to

simulate 1Hz eddy covariance, disjunct eddy covariance and REA. It was concluded that,

Chapter 4

88

when using a scroll pump is not possible for technical or practical reasons; disjunct eddy

covariance was the most reliable substitute for 10Hz eddy covariance. This method

showed the highest degree of resemblance with the 10Hz eddy covariance for all three

gases. The simulated 1Hz eddy covariance of CO2 and water vapour fluxes, showed

relatively large deviations from the 10Hz eddy covariance data, indicating a reduced

reliability of the method and no possibility to test the validity of the method at sites

without the eddy covariance set-up for CH4. Simulation of REA did show similar results

for all three gases; however, the standard deviation of the time series was significantly

higher than that of the 10Hz eddy covariance, pointing at relatively large deviations from

the 10Hz measurements at the half hourly basis. REA was therefore evaluated to generate

reliable estimates of fluxes over periods in the order of days or weeks. Importantly, the b-

value for CH4 flux measurements was the same as that for the CO2 and water vapour flux

measurements. This indicates that it will be possible to determine the b-value for CH4

measurements with REA at new and remote locations using an eddy covariance set-up

with low power requirements for CO2 or water vapour. Finally, it should be taken into

account that the ‘snap sampling’ instrument for disjunct eddy covariance and the valve

switching system for REA will introduce additional artefacts, which also require certain

amounts of power (Rinne et al., 2000; Kuhn et al., 2005; Graus et al., 2006).

Acknowledgements

This research project is performed in the framework of the European research programme

Carbo Europe (contract number GOCE-CT2003-505572) and the Dutch National

Research Programme Climate Changes Spatial Planning (www.klimaatvooruimte.nl). We

very much appreciate the help from Alex Vermeulen working at ECN for the information

and data from the Cabauw site. Also, we would very much like to thank our technical co-

operators Rob Stoevelaar and Ron Lootens.

An eddy covariance set-up for methane measurements

89

References

Allan, D. W.: Statistics of Atomic Frequency Standards. Proceedings of the Institute of

Electrical and Electronics Engineers 54(2): 221-&, 1966.

Aubinet, M., Grelle A., Ibrom, A. et al.: Estimates of the annual net carbon and water

exchange of forests: The EUROFLUX methodology. Advances in Ecological

Research, 30: 113-175, 2000.

Bear, D. S., Paul, J. B., Gupta, M. et al.: Sensitive absorption measurements in the near-

infrared region using off-axis integrated-cavity-output spectroscopy. Applied

Physics B 75, 261 - 265, 2002.

Bowling, D. R., Delany, A. C., Turnipseed, A. A. et al.: Modification of the relaxed eddy

accumulation technique to maximize measured scalar mixing ratio differences in

updrafts and downdrafts. Journal of Geophysical Research-Atmospheres 104(D8):

9121-9133, 1999.

Businger, J. A. and Oncley, S. P.: Flux Measurement with Conditional Sampling. Journal

of Atmospheric and Oceanic Technology 7(2): 349-352, 1990.

Dolman, A.J., Maximov, T. C., Moors, E. J. et al.: Net ecosystem exchange of carbon

dioxide and water of far eastern Siberian Larch (Larix cajanderii) on permafrost.

Biogeosciences, 1: 133-146, 2004.

Drösler, M., Freibauer, A., Christensen, T. R. et al.: Observations and status of peatland

greenhouse gas emissions in Europe. In: Dolman, A. J., Valentini, R. and Freibauer,

A.: Observing the continental scale greenhouse gas balance, Chapter 8. To be

published in Springer Ecological series, 2007.

Eisma, R., Vermeulen, A. T. and Kieskamp, W.M.: Determination of European methane

emissions, using concentration and isotope measurements. Environmental

Monitoring and Assessment 31: 197 – 202, 1994.

Faist, J., Capasso, F., Sivco, D. L. et al.: Quantum Cascade Laser - Temperature-

Dependence of the Performance-Characteristics and High T0 Operation. Applied

Physics Letters 65(23): 2901-2903, 1994.

Foster, P., Ramaswamy, V., Artaxo, P. et al.: Changes in Atmospheric Constituents and in

Radiative Forcing, in: Climate Change 2007: The Physical Science Basis.

Contribution of working group I to the Fourth Assessment Report of the

Intergovernmental Panel on Climate Change, edited by: Solomon, S., Qin., D.,

Manning, M. et al., Cambridge University Press, Cambridge, United Kingdom and

New York, NY, USA, 2007.

Gash, J. H. C. and Dolman, A. J.: Sonic anemometer (co)sine response and flux

measurement I. The potential for (co)sine error to affect sonic anemometer-based

flux measurements, Agricultural and Forest Meteorology 119(3-4): 195-207, 2003.

Chapter 4

90

Graus, M., Hansel, A., Wisthaler, A. et al.: A relaxed-eddy-accumulation method for the

measurement of isoprenoid canopy-fluxes using an online gas-chromatographic

technique and PTR-MS simultaneously. Atmospheric Environment 40: S43-S54,

2006.

Hargreaves, K. J., Fowler, D., Pitcairn, C. E. R. et al.: Annual methane emission from

Finnish mires estimated from eddy covariance campaign measurements. Theoretical

and Applied Climatology 70(1-4): 203-213, 2001.

Hendriks, D. M. D., van Huissteden, J., Dolman, A. J., van der Molen, M. K.: The full

greenhouse gas balance of an abandoned peat meadow. Biogeosciences 4: 411-424,

2007.

Højstrup, J.: A simple model for the adjustment of velocity spectra in unstable conditions

downstream of an abrupt change in roughness and heat flux. Boundary-Layer

Meteorology 21: 341-356 (1981).

Kaimal, J. C. and Finnigan, J. J.: Atmospheric boundary layer flows, their structure and

management, New York, Oxford University Press, 1994.

Kaimal, J. C., Wyngaard, J. C., Haugen, D. A., et al.: Turbulence structure in convective

boundary-layer. Journal of the atmospheric 33 (11): 2152-2169, 1976.

Kaimal, J. C., Izumi, Y, Wyngaard, J. C. et al.: Spectral characteristics of surface-layer

turbulence. Quarterly Journal of the Royal Meteorological Society 98 (417): 563-&,

1972.

Kormann R., Muller, H., Werle, P.: Eddy flux measurements of methane over the fen

"Murnauer Moos", 11 degrees 11 ' E, 47 degrees 39 ' N, using a fast tunable diode

laser spectrometer. Atmospheric Environment 35(14): 2533-2544, 2001.

Kroon, P.S., Hensen, A., Zahniser, M.S. et al.: Suitability of quantum cascade laser

spectrometry for CH4 and N2O eddy covariance measurements. Biogeosciences 4:

715-728, 2007.

Kuhn, U., Dindorf, T., Ammann, C. et al.: Design and field application of an automated

cartridge sampler for VOC concentration and flux measurements. Journal of

environmental monitoring 7 (6): 568-576, 2005.

Leuning, R. and Moncrieff, J.: Eddy-Covariance CO2 Flux Measurements Using Open-

Path and Closed-Path CO2 Analyzers - Corrections for Analyzer Water-Vapour

Sensitivity and Damping of Fluctuations in Air Sampling Tubes. Boundary-Layer

Meteorology 53(1-2): 63-76, 1990.

Leuning, R. and King, K. M.: Comparison of eddy-covariance measurements of CO2

fluxes by open-path and closed path CO2 analyzers. Boundary-layer meteorology 59

(3): 297-311, 1992.

Lloyd, C. R., Bessemoulin, P., Cropley, F. D. et al.: A comparison of surface fluxes at the

HAPEX-Sahel fallow bush sites, Journal of Hydrology 189(1-4): 400-425, 1997.

An eddy covariance set-up for methane measurements

91

Moore, C. J.: Frequency response corrections for eddy-correlation systems. Boundary

layer meteorology, 37 (1-2): 17-35, 1986.

Nakai, T., van der Molen, M. K., Gash, J. H. C. et al.: Correction of sonic anemometer

angle of attack errors. Agricultural and Forest Meteorology 136(1-2): 19-30, 2006.

Nelson, D. D., McManus, B., Urbanski, S. et al.: High precision measurements of

atmospheric nitrous oxide and methane using thermoelectrically cooled mid-infrared

quantum cascade lasers and detectors. Spectrochimica Acta Part a-Molecular and

Biomolecular Spectroscopy 60(14): 3325-3335, 2004.

O'Keefe, A.: Integrated cavity output analysis of ultra-weak absorption. Chemical Physics

Letters 293(5-6): 331-336, 1998.

Pattey, E., Desjardins, R. L. and Rochette, P.: Accuracy of the Relaxed Eddy-

Accumulation Technique, Evaluated Using CO2 Flux Measurements. Boundary-

Layer Meteorology 66(4): 341-355, 1993.

Rinne, H. J. I., Delany, A. C., Greenberg, J. P. et al.: A true eddy accumulation system for

trace gas fluxes using disjunct eddy sampling method. Journal of Geophysical

Research-Atmospheres 105(D20): 24791-24798, 2000.

Rinne, H. J. I., Guenther, A. B., Warneke, C. et al.: Disjunct eddy covariance technique

for trace gas flux measurements. Geophysical Research Letters 28(16): 3139-3142,

2001.

Smeets, C. J. P. P., Duynkerke, P. G., Vugts, H. F.: Turbulence characteristics of the

stable boundary layer over a mid-latitude glacier. Part I: A combination of katabatic

and large-scale forcing. Boundary layer meteorology 87 (1): 117-145, 1998.

Stull, R.B.: An Introduction to boundary layer meteorology, Dordrecht, Kluwer Academic

Publishers, Atmospheric Sciences Library, Chapter 8, 1988.

Van der Molen, M.K.: Meteorological impacts of land use change in the Maritime tropics.

Chapter 3: Data analysis, Thesis, Vrije Universiteit, 35 - 72, 2002.

Van der Molen, M. K., Gash, J. H. C., Elbers, J. A. et al.: Sonic anemometer (co)sine

response and flux measurement - II. The effect of introducing an angle of attack

dependent calibration, Agricultural and Forest Meteorology 122(1-2): 95-109, 2004.

Van der Molen, M. K., Zeeman, M. J., Lebis, J. et al.: EClog: A handheld eddy covariance

logging system. Computers and Electronics in Agriculture 51(1-2): 110-114, 2006.

Verma, S.B., Ullman, F.G., Billesbach, D. et al.: Eddy correlation measurements of

methane flux in a northern peatland ecosystem. Boundary-Layer meteorology 58:

289-304, 1992.

Werle, P., Mucke, R., and Slemr, F.: The Limits of Signal Averaging in Atmospheric

Trace-Gas Monitoring by Tunable Diode-Laser Absorption-Spectroscopy (Tdlas).

Applied Physics B-Photophysics and Laser Chemistry 57(2): 131-139, 1993.

Chapter 4

92

Wohlfahrt, G., Anfang, C., Bahn, M. et al.: Quantifying nighttime ecosystem respiration

of a meadow using eddy covariance, chambers and modelling. Agricultural and

Forest Meteorology 128(3-4): 141-162, 2005.

Zahniser, M. S., Nelson, D. D., McManus, J. B. et al.: Measurement of Trace Gas Fluxes

Using Tunable Diode-Laser Spectroscopy. Philosophical Transactions of the Royal

Society of London Series a-Mathematical Physical and Engineering Sciences

351(1696): 371-381, 1995.